def generate_new_no_bg_files(in_dir, out_dir, fg_path=None, bg_path=None):
""" Generate the new JAMS and audio files with a different background SNR
Args:
new_snr: float, Sound to noise ratio (SNR) of the background from the reference
in_dir: str, folder containing JAMS file with background SNR to be changed
out_dir: str, folder where to save the new audio and JAMS
Returns:
"""
logger = create_logger(__name__, "Desed.log")
create_folder(out_dir)
for jam_file in sorted(glob.glob(osp.join(in_dir, "*.jams"))):
jams_obj = delete_bg(jam_file)
out_jams = osp.join(out_dir, osp.basename(jam_file))
jams_obj.save(out_jams)
audiofile = osp.join(out_dir,
osp.splitext(osp.basename(jam_file))[0] + ".wav")
logger.debug(audiofile)
scaper.generate_from_jams(out_jams,
audiofile,
fg_path,
bg_path,
jams_outfile=out_jams)
def generate_audio(jam, anns, background_path, foreground_path, out_file_path):
"""The function generates an audio clip from a jam file given as input
Args
jam (JAMS): jam file
anns (Annotations): annotations of the jam file
background_path (str): path to background soundbank folder
foreground_path (str): path to foreeground soundbank folder
out_file_path (str): path to output file
Returns:
None
"""
jam.annotations.append(anns)
jam.save(out_file_path)
# generate the audio clip
scaper.generate_from_jams(
out_file_path,
audio_outfile=out_file_path.replace(".jams", ".wav"),
fg_path=foreground_path,
bg_path=background_path,
jams_outfile=out_file_path,
save_isolated_events=False,
txt_path=out_file_path.replace(".jams", ".txt"),
)
def generate_files_from_jams(list_jams, out_folder, out_folder_jams=None,
save_isolated_events=False, overwrite_exist_audio=False,
**kwargs):
""" Generate audio files from jams files generated by Scaper
Args:
list_jams: list, list of jams filepath generated by Scaper.
out_folder: str, output path to save audio files.
out_folder_jams: str, path to write the jams (could be modified by fg_path and bg_path for example),
if None, jams not saved
save_isolated_events: bool, whether or not to save isolated events in a separate folder
overwrite_exist_audio: bool, whether to regenerate existing audio files or not
kwargs: dict, scaper.generate_from_jams params (fg_path, bg_path, ...)
Returns: None
"""
logger = create_logger(__name__ + "/" + inspect.currentframe().f_code.co_name)
logger.info(f"generating audio files to {out_folder}")
create_folder(out_folder)
for n, jam_file in enumerate(list_jams):
logger.debug(jam_file)
audiofile = osp.join(out_folder, f"{osp.splitext(osp.basename(jam_file))[0]}.wav")
if not osp.exists(audiofile) or overwrite_exist_audio:
if out_folder_jams is not None:
jams_outfile = osp.join(out_folder_jams, osp.basename(jam_file))
else:
jams_outfile = None
generate_from_jams(jam_file, audiofile, jams_outfile=jams_outfile,
save_isolated_events=save_isolated_events,
**kwargs)
if n % 200 == 0:
logger.info(f"generating {n} / {len(list_jams)} files (updated every 200)")
logger.info("Done")
def generate_files(list_jams, outfolder, fg_path=None, bg_path=None, overwrite_jams=False):
n = 0
for jam_file in list_jams:
audiofile = osp.join(outfolder, f"{osp.splitext(osp.basename(jam_file))[0]}.wav")
if overwrite_jams:
jams_outfile = jam_file
else:
jams_outfile = None
generate_from_jams(jam_file, audiofile, fg_path=fg_path, bg_path=bg_path, jams_outfile=jams_outfile)
n += 1
if n % 500 == 0:
print(f"generating {n} / {len(list_jams)} files")
print("Done")
def generate_new_fg_onset_files(added_value, in_dir, out_dir):
""" Generate the new JAMS and audio files adding a value to forground onsets
Args:
added_value: float, value in seconds, value to be added to previous onset
in_dir: str, folder containing JAMS file with background SNR to be changed
out_dir: str, folder where to save the new audio and JAMS
Returns:
"""
for jam_file in sorted(glob.glob(os.path.join(in_dir, "*.jams"))):
jams_obj = modify_fg_onset(added_value, jam_file)
out_jams = osp.join(out_dir, os.path.basename(jam_file))
jams_obj.save(out_jams)
audiofile = os.path.join(out_dir, osp.splitext(osp.basename(jam_file))[0] + ".wav")
print(audiofile)
scaper.generate_from_jams(out_jams, audiofile)
def generate_new_bg_snr_files(new_snr, in_dir, out_dir):
""" Generate the new JAMS and audio files with a different background SNR
Args:
new_snr: float, Sound to noise ratio (SNR) of the background from the reference
in_dir: str, folder containing JAMS file with background SNR to be changed
out_dir: str, folder where to save the new audio and JAMS
Returns:
"""
for jam_file in sorted(glob.glob(os.path.join(in_dir, "*.jams"))):
jams_obj = modify_bg_snr(new_snr, jam_file)
out_jams = osp.join(out_dir, os.path.basename(jam_file))
jams_obj.save(out_jams)
audiofile = os.path.join(
out_dir,
osp.splitext(osp.basename(jam_file))[0] + ".wav")
print(audiofile)
scaper.generate_from_jams(out_jams, audiofile)
def synthesize_one_mixture(jams_file):
wav_file = jams_file.replace('.jams', '.wav')
generate_from_jams(jams_file, wav_file, save_sources=True)
def test_generate_from_jams(atol=1e-5, rtol=1e-8):
# Test for invalid jams: no annotations
tmpfiles = []
with _close_temp_files(tmpfiles):
jam = jams.JAMS()
jam.file_metadata.duration = 10
jam_file = tempfile.NamedTemporaryFile(suffix='.jams', delete=True)
gen_file = tempfile.NamedTemporaryFile(suffix='.jams', delete=True)
jam.save(jam_file.name)
pytest.raises(ScaperError, scaper.generate_from_jams, jam_file.name,
gen_file.name)
# Test for valid jams files
tmpfiles = []
with _close_temp_files(tmpfiles):
# Create all necessary temp files
orig_wav_file = tempfile.NamedTemporaryFile(suffix='.wav', delete=True)
orig_jam_file = tempfile.NamedTemporaryFile(suffix='.jams',
delete=True)
gen_wav_file = tempfile.NamedTemporaryFile(suffix='.wav', delete=True)
gen_jam_file = tempfile.NamedTemporaryFile(suffix='.jams', delete=True)
tmpfiles.append(orig_wav_file)
tmpfiles.append(orig_jam_file)
tmpfiles.append(gen_wav_file)
tmpfiles.append(gen_jam_file)
# --- Define scaper --- *
sc = scaper.Scaper(10, FG_PATH, BG_PATH)
sc.protected_labels = []
sc.ref_db = -50
sc.add_background(label=('choose', []),
source_file=('choose', []),
source_time=('const', 0))
# Add 5 events
for _ in range(5):
sc.add_event(label=('choose', []),
source_file=('choose', []),
source_time=('const', 0),
event_time=('uniform', 0, 9),
event_duration=('choose', [1, 2, 3]),
snr=('uniform', 10, 20),
pitch_shift=('uniform', -1, 1),
time_stretch=('uniform', 0.8, 1.2))
# generate, then generate from the jams and compare audio files
# repeat 5 time
for _ in range(5):
sc.generate(orig_wav_file.name,
orig_jam_file.name,
disable_instantiation_warnings=True)
scaper.generate_from_jams(orig_jam_file.name, gen_wav_file.name)
# validate audio
orig_wav, sr = soundfile.read(orig_wav_file.name)
gen_wav, sr = soundfile.read(gen_wav_file.name)
assert np.allclose(gen_wav, orig_wav, atol=atol, rtol=rtol)
# Now add in trimming!
for _ in range(5):
sc.generate(orig_wav_file.name,
orig_jam_file.name,
disable_instantiation_warnings=True)
scaper.trim(orig_wav_file.name, orig_jam_file.name,
orig_wav_file.name, orig_jam_file.name,
np.random.uniform(0, 5), np.random.uniform(5, 10))
scaper.generate_from_jams(orig_jam_file.name, gen_wav_file.name)
# validate audio
orig_wav, sr = soundfile.read(orig_wav_file.name)
gen_wav, sr = soundfile.read(gen_wav_file.name)
assert np.allclose(gen_wav, orig_wav, atol=atol, rtol=rtol)
# Double trimming
for _ in range(2):
sc.generate(orig_wav_file.name,
orig_jam_file.name,
disable_instantiation_warnings=True)
scaper.trim(orig_wav_file.name, orig_jam_file.name,
orig_wav_file.name, orig_jam_file.name,
np.random.uniform(0, 2), np.random.uniform(8, 10))
scaper.trim(orig_wav_file.name, orig_jam_file.name,
orig_wav_file.name, orig_jam_file.name,
np.random.uniform(0, 2), np.random.uniform(4, 6))
scaper.generate_from_jams(orig_jam_file.name, gen_wav_file.name)
# Tripple trimming
for _ in range(2):
sc.generate(orig_wav_file.name,
orig_jam_file.name,
disable_instantiation_warnings=True)
scaper.trim(orig_wav_file.name, orig_jam_file.name,
orig_wav_file.name, orig_jam_file.name,
np.random.uniform(0, 2), np.random.uniform(8, 10))
scaper.trim(orig_wav_file.name, orig_jam_file.name,
orig_wav_file.name, orig_jam_file.name,
np.random.uniform(0, 1), np.random.uniform(5, 6))
scaper.trim(orig_wav_file.name, orig_jam_file.name,
orig_wav_file.name, orig_jam_file.name,
np.random.uniform(0, 1), np.random.uniform(3, 4))
scaper.generate_from_jams(orig_jam_file.name, gen_wav_file.name)
# validate audio
orig_wav, sr = soundfile.read(orig_wav_file.name)
gen_wav, sr = soundfile.read(gen_wav_file.name)
assert np.allclose(gen_wav, orig_wav, atol=atol, rtol=rtol)
# Test with new FG and BG paths
for _ in range(5):
sc.generate(orig_wav_file.name,
orig_jam_file.name,
disable_instantiation_warnings=True)
scaper.generate_from_jams(orig_jam_file.name,
gen_wav_file.name,
fg_path=ALT_FG_PATH,
bg_path=ALT_BG_PATH)
# validate audio
orig_wav, sr = soundfile.read(orig_wav_file.name)
gen_wav, sr = soundfile.read(gen_wav_file.name)
assert np.allclose(gen_wav, orig_wav, atol=atol, rtol=rtol)
# Ensure jam file saved correctly
scaper.generate_from_jams(orig_jam_file.name,
gen_wav_file.name,
jams_outfile=gen_jam_file.name)
orig_jam = jams.load(orig_jam_file.name)
gen_jam = jams.load(gen_jam_file.name)
assert orig_jam == gen_jam
def test_generate_from_jams(atol=1e-5, rtol=1e-8):
# Test for invalid jams: no annotations
tmpfiles = []
with _close_temp_files(tmpfiles):
jam = jams.JAMS()
jam.file_metadata.duration = 10
jam_file = tempfile.NamedTemporaryFile(suffix='.jams', delete=True)
gen_file = tempfile.NamedTemporaryFile(suffix='.jams', delete=True)
jam.save(jam_file.name)
pytest.raises(ScaperError, scaper.generate_from_jams, jam_file.name,
gen_file.name)
# Test for valid jams files
tmpfiles = []
with _close_temp_files(tmpfiles):
# Create all necessary temp files
orig_wav_file = tempfile.NamedTemporaryFile(suffix='.wav', delete=True)
orig_jam_file = tempfile.NamedTemporaryFile(suffix='.jams', delete=True)
gen_wav_file = tempfile.NamedTemporaryFile(suffix='.wav', delete=True)
gen_jam_file = tempfile.NamedTemporaryFile(suffix='.jams', delete=True)
tmpfiles.append(orig_wav_file)
tmpfiles.append(orig_jam_file)
tmpfiles.append(gen_wav_file)
tmpfiles.append(gen_jam_file)
# --- Define scaper --- *
sc = scaper.Scaper(10, FG_PATH, BG_PATH)
sc.protected_labels = []
sc.ref_db = -50
sc.add_background(label=('choose', []),
source_file=('choose', []),
source_time=('const', 0))
# Add 5 events
for _ in range(5):
sc.add_event(label=('choose', []),
source_file=('choose', []),
source_time=('const', 0),
event_time=('uniform', 0, 9),
event_duration=('choose', [1, 2, 3]),
snr=('uniform', 10, 20),
pitch_shift=('uniform', -1, 1),
time_stretch=('uniform', 0.8, 1.2))
# generate, then generate from the jams and compare audio files
# repeat 5 time
for _ in range(5):
sc.generate(orig_wav_file.name, orig_jam_file.name,
disable_instantiation_warnings=True)
scaper.generate_from_jams(orig_jam_file.name, gen_wav_file.name)
# validate audio
orig_wav, sr = soundfile.read(orig_wav_file.name)
gen_wav, sr = soundfile.read(gen_wav_file.name)
assert np.allclose(gen_wav, orig_wav, atol=atol, rtol=rtol)
# Now add in trimming!
for _ in range(5):
sc.generate(orig_wav_file.name, orig_jam_file.name,
disable_instantiation_warnings=True)
scaper.trim(orig_wav_file.name, orig_jam_file.name,
orig_wav_file.name, orig_jam_file.name,
np.random.uniform(0, 5), np.random.uniform(5, 10))
scaper.generate_from_jams(orig_jam_file.name, gen_wav_file.name)
# validate audio
orig_wav, sr = soundfile.read(orig_wav_file.name)
gen_wav, sr = soundfile.read(gen_wav_file.name)
assert np.allclose(gen_wav, orig_wav, atol=atol, rtol=rtol)
# Double trimming
for _ in range(2):
sc.generate(orig_wav_file.name, orig_jam_file.name,
disable_instantiation_warnings=True)
scaper.trim(orig_wav_file.name, orig_jam_file.name,
orig_wav_file.name, orig_jam_file.name,
np.random.uniform(0, 2), np.random.uniform(8, 10))
scaper.trim(orig_wav_file.name, orig_jam_file.name,
orig_wav_file.name, orig_jam_file.name,
np.random.uniform(0, 2), np.random.uniform(4, 6))
scaper.generate_from_jams(orig_jam_file.name, gen_wav_file.name)
# Tripple trimming
for _ in range(2):
sc.generate(orig_wav_file.name, orig_jam_file.name,
disable_instantiation_warnings=True)
scaper.trim(orig_wav_file.name, orig_jam_file.name,
orig_wav_file.name, orig_jam_file.name,
np.random.uniform(0, 2), np.random.uniform(8, 10))
scaper.trim(orig_wav_file.name, orig_jam_file.name,
orig_wav_file.name, orig_jam_file.name,
np.random.uniform(0, 1), np.random.uniform(5, 6))
scaper.trim(orig_wav_file.name, orig_jam_file.name,
orig_wav_file.name, orig_jam_file.name,
np.random.uniform(0, 1), np.random.uniform(3, 4))
scaper.generate_from_jams(orig_jam_file.name, gen_wav_file.name)
# validate audio
orig_wav, sr = soundfile.read(orig_wav_file.name)
gen_wav, sr = soundfile.read(gen_wav_file.name)
assert np.allclose(gen_wav, orig_wav, atol=atol, rtol=rtol)
# Test with new FG and BG paths
for _ in range(5):
sc.generate(orig_wav_file.name, orig_jam_file.name,
disable_instantiation_warnings=True)
scaper.generate_from_jams(orig_jam_file.name, gen_wav_file.name,
fg_path=ALT_FG_PATH, bg_path=ALT_BG_PATH)
# validate audio
orig_wav, sr = soundfile.read(orig_wav_file.name)
gen_wav, sr = soundfile.read(gen_wav_file.name)
assert np.allclose(gen_wav, orig_wav, atol=atol, rtol=rtol)
# Ensure jam file saved correctly
scaper.generate_from_jams(orig_jam_file.name, gen_wav_file.name,
jams_outfile=gen_jam_file.name)
orig_jam = jams.load(orig_jam_file.name)
gen_jam = jams.load(gen_jam_file.name)
assert orig_jam == gen_jam